Very early in my experimentation with some of the GPT family, one of the first things I noticed was that at no point did the model ask questions of its own, except (perhaps - I don't quite recall if I'm the one hallucinating here ) to indicate that further clarification of a previous prompt would improve a response.I haven't touched them in quite a long time and I don't know if they now do this, but it struck me that their lack of (even simulated) curiosity was a major void in the popular dialogue about whether they are or are not "intelligent."
No comments yet.